How the archive explorer works

This guide makes the project’s concepts and workflow more understandable to human reviewers and contributors.

Four interactive, source-linked diagrams explain the project’s evidence, attribution and review logic. Open a diagram, select a node for its detail and source references, and use its Live/Still control to pause movement.

Read these as a documented workflow at private source revision 42aade8, not proof that every authorship claim is resolved or that all human/accessibility review has happened. Later project changes are outside that frozen source scope; consult the current methods and work log alongside these diagrams.

Start with the whole project

Follow the XML snapshot, reviewed records, historical role evidence and saved analysis into the allowlisted public site.

Text equivalent: steps and relationships

Connections:

  • WordPress XML → Extract + review (saved XML).
  • Extract + review → Archive ledger (record fields).
  • Role evidence → Match role history (role claims).
  • Archive ledger → Match role history (credit + date).
  • Archive ledger → Build explorer (archive records).
  • Saved text results → Build explorer (saved outputs).
  • Match role history → Build explorer (role context).
  • Build explorer → Public release (public allowlist).

How the text is used

Compare the historical cleaned-corpus/MALLET path with literal sentence extraction and the selected or broader excerpt collections.

Text equivalent: steps and relationships

Connections:

  • Post text → NLTK cleaning (saved text).
  • Audited metadata → NLTK cleaning (metadata).
  • NLTK cleaning → Cleaned corpus (cleaned forms).
  • Cleaned corpus → MALLET fit (model words).
  • MALLET fit → Topic weights (saved weights).
  • Post text → Literal spans (literal text).
  • Audited metadata → Literal spans (eligible posts).
  • Literal spans → Excerpt pool (filter spans).
  • Excerpt pool → Select voices (reviewed IDs).
  • Select voices → Excerpt reader (selected voices).
  • Excerpt pool → Excerpt reader (broader option).

Who wrote it, and who was a Fellow?

See why writer credits, posting accounts, curators, identity mappings and dated appointments remain separate claims.

Text equivalent: steps and relationships

Connections:

  • Inspect the post → Check writer evidence.
  • Check writer evidence → Supported writer credit (Supported).
  • Check writer evidence → Provisional credit (Unresolved).
  • Supported writer credit → Keep original names.
  • Provisional credit → Keep original names.
  • Keep original names → Match credit + date.
  • Saved role history → Match credit + date (Role evidence).
  • Match credit + date → Show separate claims.

How changes are reviewed and published

Distinguish human direction, automated checks, AI simulations, process disclosure and specific publication authorization.

Text equivalent: steps and relationships

Connections:

  • Human sets scope → Inspect + implement.
  • Inspect + implement → Run relevant checks.
  • Inspect + implement → AI review simulation (If performed).
  • Human sets scope → Record human input (If supplied).
  • Run relevant checks → Inspect + implement (Repair failures).
  • Run relevant checks → Disclose the process (Checks pass).
  • Record human input → Disclose the process.
  • AI review simulation → Disclose the process.
  • Disclose the process → Release authorized?.
  • Release authorized? → Allowlisted release (Authorized).
  • Release authorized? → Hold or revise (Not authorized).